A new paper entitled "Adverbs, Surprisingly" by Dmitry Nikolaev, Collin F. Baker, Miriam R.L. Petruck, and Sebastian Padó was presented as a poster at the StarSEM 2023 workshop (July 13-14), colocated with the annual ACL conference in Toronto, Canada. Dmitry Nikolaev, a postdoc at the Institute for Natural Language Processing, University of Stuttgart, gave the presentation; the paper is available on the ACL Anthology.
This paper begins with the premise that adverbs are generally neglected in computational linguistics. This view is based on two analyses: a literature review and the results of using a novel adverb dataset to probe a state-of-the-art language model, which reveals systematic gaps in standard accounts of adverb meaning. We suggest that using Frame Semantics to characterize word meaning, as in FrameNet, is a promising way to analyze adverbs, given its ability to represent ambiguity, semantic roles, and null instantiation.